MongoDB Tit Bits

While I am getting started with MongoDB for my new project. Here are the few quick notes - based on my readings so far:

One Key Takeaway:

Data accessed together should be stored together

WiredTiger:

• WiredTiger is the default storage engine for mongo db
• It stores documents and indexes on disk
• In memory cache stores some doc and frequency used index - working set
  • 50% of (RAM -1 GB) Or, 256MB

Massive Arrays:

• Max Document size is 16MB
• Index performance on arrays decreases as array size increases
• Extended Reference pattern: where we duplicate some and not all data

Index:

• Each index is atleast 8KB
• Index take up storage - One File for each collection and one file for each Index (WiredTiger impl)
• Write performance as index needs to updated
• Limit each collection to 50 Index max
• Do: Add index for frequently supported queries - improves read performance
• Don’t: Create unnecessary indexes - reduce performance and takes up space

Bloated Documents:

• Do: Data accessed together should be stored together
• Don’t: bloat your document with related data that is not accessed to gether
• Data that is related to each other should NOT necessarily stored together
• Remove bloat from frequently used documents - it can be inmemory wire tiger cache
• Data Duplication is OK (depends!)
  • Summary document and Details document - ref via links

Case Incentive Query:

• $regex queries are case insensitive but not performant
• Non-$regex queries are case sensitive
• Collation:
  • Language specific rules for MongoDB for string comparison
  • Strength ranges from 1-5
  • Strength 1-2 will give you case insensitive
  • Query: {“first_name”: {$regex: /Jacob/i }} - case insensitive
  • Query: {“first_name”: “Jacob”} - case sensitive

Accessing - Separate data together:

• $lookup:
  • Is used to join data from more than one collection
  • Great for rarely used queries or analytical queries (batch run overnight)
  • Very slow and resource intensive

Upsert:

• Upsert is a combination of update and insert. Upsert performs two functions:
  • Update data if there is a matching document.
  • Insert a new document in case there is no document matches the query criteria.
• Personally: I found this very interesting, as for my use cases - I dont need to track if the document already exist.

I understand - the blog looks a bit navive (read, nothing interseting), sry I am just getting started. BTW, I plan to write a follow blog - deep dive on my specific use cases and data modelling.

Reference:

• MongoDB Channel
• MongoDB Tutorial